Introduction

Study 1 has shown that there is a correlation between sex hormones and spatial navigation strategy, such that higher estradiol is positively associated with higher use of shortcuts in the DSP task (higher solution index). Higher FSH, on the other hand, was marginally negatively associated with shortcut use (lower solution index) using Spearman rank correlations.

Here, I determine whether there are associations between T2 high-resolution hippocampal subfield volume and spatial navigation behaviors, and sex hormones and subfield volume. Later, I determine how subfield volume mediates the relationship between sex hormones and behavior for navigation strategy.

Prior exploratory analysis for relationships between sex hormones and subfield volume reveal certain significant relationships (correlation_greymatter_hormones_corrected.Rmd).

Here, I will explore whether there are any relationships between spatial navigation behaviors and T2 subfield volumes. Next, I plan to apply causal mediation analysis to investigate whether the effect of sex hormones (IV) on spatial behavior (DV) is mediated by participants’ subfield volume (mediator, T2 subfield).

Data Analysis

I cleaned some of the data prior to analyses. I had to:

  1. Filter for midlife participants that have T2 subfield data (subject 333 and over)
  2. Select sex hormone measures from when scan was conducted
  3. Correct for subfield volumes using TIV, which is a combination of GMV, WMV, CSF. Divide the volumes by TIV variable (t2hipp_vol_tiv)
  4. Average left and right volumes to get an average measure for each subfield and brain region

Correlations for Behavior and T2 Subfields

Behavioral Distributions

Overall

From prior analyses, we know that the subfield volumes are generally normally distributed, with the exception of CA23 in men and SUB in women and overall.

We can check the distribution of the SNAG dependent variables

LOOP dependent variables appear normally distributed, but MAZE and DSP variables have slight skews. I applied log and square root transformations to these variables in MAZE and DSP to see whether the observed distribution looks normal.

We can check Shapiro-Wilk’s to see if normality is violated for raw values and after transformation using log and square root.

Shapiro-Wilk’s test reveal that LOOP dependent variables are normally distributed, but MAZE and DSP dependent variables are not even after log and square root transformations. Thus, all correlations with subfields for these DV will be spearman rank correlations.

By Women

Like the overall distributions, LOOP DVs look normal but MAZE and DSP variables appear skewed.

We can check Shapiro-Wilk’s to see if normality is violated for raw values and after transformation using log and square root.

LOOP and DSP wayfinding succss was not significantly different than normal. DSP solution index was still skewed after transformations, so correlations with MAZE varaibles and DSP solution index for women will be using Spearman rank correlation.

By Men

Like the overall distributions, LOOP DVs look normal but MAZE and DSP variables appear skewed.

LOOP and DSP and MAZE moves variables were not significantly different than normal, but MAZE accuracy was. Despite transformations, MAZE accuracy was still pretty skewed so Spearman rank correlations will be applied for associations with subfields for men here.

LOOP

Position Error

CA23 men and SUB overall and women were Spearman rank correlations. All others are Pearson correlations.

Degrees Traveled

CA23 men and SUB overall and women were Spearman rank correlations. All others are Pearson correlations.

MAZE

Wayfinding Success

All Spearman rank correlations.

Moves

All Spearman rank correlations.

DSP

Wayfinding Success

All Spearman rank correlations.

Solution Index

All Spearman rank correlations.

Summary

  • From previous analyses, CA23 in men and SUB overall and in women were not normally distributed. All correlations comapring these specific subfields were using Spearman rank correlations.

  • Overall with women and men, LOOP variables were normally distributed, but MAZE and DSP variables were not. Among women, MAZE variables and DSP solution index were not normally distributed. Among mwn, MAZE accuracy was not normally distributed. LOOP mainly used Pearson correlations (other than at CA23 and SUB), while all correlations with MAZE and DSP variables were Spearman rank.

  • For LOOP, position error was negatively correlated with PHC overall (p = 0.038, Pearson) and among women (p = 0.011, Pearson). Position error was also negatively correlated with SUB in men (p = 0.047, Spearman) and was marginally negatively correlated with ERC in men (p = 0.073, Pearson).

  • For MAZE, there were no significant associations between subfields and accuracy, or between subfields and number of moves made.

  • For DSP, there were no significant associations between subfields and wayfinding success. Solution index was marginally negatively associated with DG in men (p = 0.056, Spearman)

TBD